The PCA is a very helpful tool in multivariate analysis of real data. Correlations are difficult to avoid. Hopefully the PCA alone provides some insight, such as assets and membership possibly the expression of some latent variable.

I assume that by "although factor 3 provides the lowest explanatory value (at ~12%)" you mean 12% of the variance in the PCA. You are correct about factor 3. The fact that factor 1 represents the most variation in the original variables in no way implies that it is the most important predictor. Using all three factors in the model provides a strong analysis for separating and testing the effects.

Learn it once, use it forever!